As is done everywhere else.
From what I can gather from make -C tests/check V=1 $(GST_PLUGINS_BAD_CFLAGS) is
required in order to find in-tree headers as well as srcdir != builddir
configurations.
With e38af23044 returning the correct contexts,
gst_gl_display_add_context() was susceptible to causing infinte loops when
adding the same GstGLContext more than once. Fix and add a test for
gst_gl_display_add_context().
Fixes glvideomixer gst-validate tests.
When the start_type is GST_SEEK_TYPE_NONE for a forward seek
(or stop_type for a reverse) is not set on a snap seeking operation,
the element should use the current position and then snap as requested.
Also fixes uninitialized variable complaint by clang about
'ts' variable.
All hlsdemux tests create a GstStructure called "state" that can be used
by test cases to store information during a test. The name of this
structure is arbitrary. When the code was written, the intention was
to use the name of the test, to aid debugging. However, during
development this was lost, so that the state GstStructure is always
given the name "setup_test_variables".
This commit changes this so that the name of the test is used.
https://bugzilla.gnome.org/show_bug.cgi?id=762684
The default one is 6 minutes, the test was using 5 minutes so just
resort to using the default.
For the non-valgrind test also use the default 20 secs instead of
reducing it to 6s. No real reason to set a custom value here.
When caps are already negotiated it should be possible to
select formats other than the one that was negotiated. If downstream
allows alpha video caps and it has already negotiated to a non-alpha
format, caps queries should still return the alpha caps as a possible
format as caps renegotiation can happen.
Includes tests (for compositor) to check that caps queries done after
a caps has been negotiated returns complete results
https://bugzilla.gnome.org/show_bug.cgi?id=757610
The function actually returns the segment availability start time (as defined by the standard).
That is at the end of the segment, but it is called availability start time.
Availability end time is something else (the time when the segment is no longer
available on the server). The function name was misleading.
https://bugzilla.gnome.org/show_bug.cgi?id=757655
The demux_sent_eos callback is unused in tests. It was also registered on
a wrong pad, so it actually triggered when demux received eos from a
fragment download.
https://bugzilla.gnome.org/show_bug.cgi?id=760328
Adds unit tests similar to the ones that we have for DASH and HLS.
Tests:
* manifest parsing finishes successfully
* some queries (duration, seekable, latency)
* seeking with various values and flags
Similar to HLS but DASH has the extra issue that it can have
multiple streams so snapping can be tricky as streams usually
won't be aligned.
For now, those tests handle the case of only having a single
stream.
https://bugzilla.gnome.org/show_bug.cgi?id=759158
Handling the ghostpad and its internal pad was causing more issues
than helping because of their coupled activation/deactivation
actions.
As we have to install custom chain,event and query functions it is
better to use a floating sink pad internally in the demuxer and just
use those pad functions to push through a standard pad in the demuxer
https://bugzilla.gnome.org/show_bug.cgi?id=757951
Some operations are unnecessary when running with only a single GL
context.
e.g. glFlush when setting a fence object as the flush happens on wait.
API: gst_gl_context_is_shared
In very few cases the simple version was actually needed and having the
parameters hidden by a _full() version caused application that actually needed
it to not use it.
The URI attribute from the EXT-X-KEY tag and the URI attribute from the
EXT-X-I-FRAMES-ONLY tag are both quoted-string attibutes that have their
quotation marks removed during parsing. The CODECS attribute of the
EXT-X-STREAM-INF is also a quoted-string attribute, but this attribute
was not being un-quoted.
This commit changes the parser to always unquote all quoted-string
attributes and adjusts the unit tests to this new bevahiour for the
CODECS attribute.
An additional test is added to check that parsing of all of the fields
in the EXT-X-STREAM tag is correct, including those that contain comma
characters.
https://bugzilla.gnome.org/show_bug.cgi?id=758384
Using the new GstAdaptiveDemux test framework, add tests that
exercise hlsdemux. The following tests are added:
simpleTest
A simple playlist that contains some media URLs
testMediaPlaylist
A master playlist with a variant playlist that contains media URLs
testMediaPlaylistNotFound
A master playlist that points to a missing variant playlist
testFragmentNotFound
A master playlist with a variant playlist that contains media URLs
There is a missing media file referenced from the variant playlist.
testFragmentDownloadError
A master playlist with a variant playlist that contains media URLs
During the download of one media file, the test simulates the network
connection being dropped.
testSeek
A simple test of trying to perform a seek on an HLS stream.
To allow code from dash_demux.c to be used by other elements
that are based upon GstAdaptiveDemux, the code has been
refactored into four new files:
adaptive_demux_engine.[ch]
adaptive_demux_common.[ch]
The code in adaptive_demux_engine.c provides a generic
test engine for elements based upon GstAdaptiveDemux.
The code in adaptive_demux_common.c provides a set
of utility functions that are common between the tests
for hlsdemux and dashdemux.
As part of the refactoring, variables in structures were
renamed from using camelCase to underscore_case to match other
GStreamer source code.
The fake_http_src was renamed test_http_src and changed to use
callbacks to provide input data and error conditions. Rather than
using an array of input data that tries to encode all the
possible use cases for the GstTestHTTPSrc element, use a struct of
callbacks.
Users of this element are obliged to implement at least the src_start
callback, which provides a way to link from a URI to the settings
for that URI.
gst_mpdparser_parse_utctiming_node does not validate the parsed values completely. The following scenarios are incorrectly accepted:
- elements with no schemeIdUri property should be rejected
- elements with unrecognized UTCTiming scheme should be rejected
- elements with empty values should be rejected
The last one triggers a division by 0 in gst_dash_demux_poll_clock_drift:
clock_drift->selected_url = clock_drift->selected_url % g_strv_length (urls);
because it urls is a valid pointer to an empty array.
https://bugzilla.gnome.org/show_bug.cgi?id=759547
The base class is useful for having multiple backing memory types other
than the default. e.g. IOSurface, EGLImage, dmabuf?
The PBO transfer logic is now inside GstGLMemoryPBO which uses GstGLBuffer
to manage the PBO memory.
This also moves the format utility functions into their own file.
The videoframe-audiolevel element acts like a synchronized audio/video "level"
element. For each video frame, it posts a level-style message containing the
RMS value of the corresponding audio frames. This element needs both video and
audio to pass through it. Furthermore, it needs a queue after its video
source.
https://bugzilla.gnome.org/show_bug.cgi?id=748259
Failure by this commit 2dfa548f36, which is
to append hooks instead of prepend.
Because of this change, aggretated_cb is not called and leads to failure.
And correct to check flush stop value instead of flush start value
https://bugzilla.gnome.org/show_bug.cgi?id=757801
This no longer does anything, and it was marked as CONSTRUCT_ONLY
which means someone would really have to go out of their way to
be able to set this, which would only be done in very custom
scenarios, if ever, and those will likely target a specific
version of GStreamer then, so probably not much point keeping
it deprecated for a while before removing it.
The Onvif Streaming Specification specifies that the NTP timestamps
in the Onvif extension header indicaes the absolute UTC time associated
with the access unit. But by using running time we can not achieve that,
since a frame's running time depends on the played interval, whether a
non-flushing is done, etc. Instead we have to use the stream time.
https://bugzilla.gnome.org/show_bug.cgi?id=757688
It is now possible to update the currently used ntp-offset with a
custom serialized downstream event. The element will read the ntp-offset
property when doing the state transition from READY to PAUSED and
use that offset until it receives a "GstNtpOffset" event, which also
has a "ntp-offset" attribute in that it's structure. In case the
property is not set and no event has been received, the element will
guess the npt-offset with help of the clock. If no clock can be
retrieved, the element will error out and stop the data flow.
The same event is also used for updating the D/E-bits in the RTP
extension header. The discont flag in a buffer can be set whenver a
live/network source looses a frame, but that is not the type of
discontinuity that the onvif extension header should reflect. The
header is mainly used for playback of a track concept, in which
gaps can be present, and it's those kind of gaps that should be
highlighted with the D- and E-bits.
https://bugzilla.gnome.org/show_bug.cgi?id=757688
If a buffer or a buffer list is cached, no events serialized with the
data stream should get through. The cached buffers and events should
be purged when we stop flushing.
https://bugzilla.gnome.org/show_bug.cgi?id=757688
Split the unit tests for rtponviftimestamp and rtponvifparse
elements in separate files.
Setup and cleanup the element and pads in fixures. Make the tests work
with CK_FORK=no as well, by cleaning up the 'buffers' list when needed.
Make unit tests work when run in valgrind by unreffing all buffers,
and by not allocating any payload in RTP buffers. Since we're not
doing anything with the payload part, but we're memcmp-aring the
complete buffer memory, valgrind complained about non-initialized
memory being used.
https://bugzilla.gnome.org/show_bug.cgi?id=757688
The standard does not seem to make any particular explicit not
implicit reference to the signedness of durations, and the code
does not rely on such, nor on the negativity of the -1 value
that's used as a placeholder when a duration property is not
present in the XML.
https://bugzilla.gnome.org/show_bug.cgi?id=750847
According to the standard:
"SegmentBase, SegmentTemplate and SegmentList shall inherit
attributes and elements from the same element on a higher level.
If the same attribute or element is present on both levels,
the one on the lower level shall take precedence over the one
on the higher level."
gst_mpdparser_parse_segment_list_node will now discard any inherited
segment URLs if the parsed element defines some too.
https://bugzilla.gnome.org/show_bug.cgi?id=751832
Add the various tokens/strings for the differnet texture types (2D, rect, oes)
Changes the GLmemory api to include the GstGLTextureTarget in all relevant
functions.
Update the relevant caps/templates for 2D only textures.
Created a unit test for dashdemux. It relies on a fake SOUP HTTP src plugin
that will feed data to dashdemux. The test controls the data to be
generated and checks the correct data was received for each expected
stream.
https://bugzilla.gnome.org/show_bug.cgi?id=756322
If MPD@suggestedPresentationDelay is not present in the manifest,
dashdemux selects the fragment closest to the most recently generated
fragment. This causes a playback issue because this choice does not allow
the DASH client to build up any buffer of downloaded fragments without
pausing playback. This is because by definition new fragments appear on
the server in real-time (e.g. if segment duration is 4 seconds, a new
fragment will appear on the server every 4 seconds). If the starting
playback position was n*segmentDuration seconds behind "now", the DASH
client could download up to 'n' fragments faster than realtime before it
reached the point where it needed to wait for fragments to appear on the
server.
The MPD@suggestedPresentationDelay attribute allows a content publisher
to provide a suggested starting position that is behind the current
"live" position.
If the MPD@suggestedPresentationDelay attribute is not present, provide
a suitable default value as a property of the dashdemux element. To
allow the default presentation delay to be specified either using
fragments or seconds, the property is a string that contains a number
and a unit (e.g. "10 seconds", "4 fragments", "2500ms").
Corrected the parsing of a segment template string.
Added unit tests to test the segment template parsing.
All reported problems are now correctly handled.
https://bugzilla.gnome.org/show_bug.cgi?id=751735
When building the media segment list using a SegmentList node, the
gst_mpd_client_setup_representation function will iterate through the
list of S nodes and will expect to find a matching SegmentUrl node. If
one does not exist, the code made an illegal memory access.
https://bugzilla.gnome.org/show_bug.cgi?id=752496
A GstGLShader is now simply a collection of stages that are
compiled and linked together into a program. The uniform/attribute
interface has remained the same.
While creating caps in audiointerleave tests, bitmask is being set as 0x9
This is resulting in segmentation fault. Fix the same by typecasting to guint64
https://bugzilla.gnome.org/show_bug.cgi?id=755840
The obscured check in compositor was using the dimensions of the pad to clamp
the h/w of the pad instead of the output resolution, and was doing an incorrect
calculation to do so. Fix that by simplifying the whole calculation by using
corner coordinates. Also add a test for this bug which fell through the cracks,
and just skip all the obscured tests if the pad's alpha is 0.0.
https://bugzilla.gnome.org/show_bug.cgi?id=754107
The test_playlist_with_doubles_duration() test fails on some platforms
due to rounding errors that occur when m3u8.c converts from the floating
point value in the HLS manifest to a GstClockTime.
Using assert_equals_float() fixes this because this function handles
the rounding error issues by accepting almost equal.
https://bugzilla.gnome.org/show_bug.cgi?id=753881
Previous patch did not handle the case where an encoding (e.g. UTF-8) is
specified in the <xml ?> element. Added an extra test for with and without
encoding.
https://bugzilla.gnome.org/show_bug.cgi?id=753813
When running on an STB, the function
gst_mpdparser_get_xml_node_as_string causes a segmentation fault. This
code works correctly on a Linux desktop.
Looking at the libxml documentation, the xmlNodeDump is deprecated.
Replacing the use of xmlNodeDump with xmlNodeDumpOutput fixes the
segfault on the STB and removes the use of the deprecated function.
Unless the DASH client can compensate for the difference between its
clock and the clock used by the server, the client might request
fragments that either not yet on the server or fragments that have
already been expired from the server. This is an issue because these
requests can propagate all the way back to the origin
ISO/IEC 23009-1:2014/Amd 1 [PDAM1] defines a new UTCTiming element to allow
a DASH client to track the clock used by the server generating the
DASH stream. Multiple UTCTiming elements might be present, to indicate
support for multiple methods of UTC time gathering. Each element can
contain a white space separated list of URLs that can be contacted
to discover the UTC time from the server's perspective.
This commit provides parsing of UTCTiming elements, unit tests of this
parsing and a function to poll a time server. This function
supports the following methods:
urn:mpeg:dash:utc:ntp:2014
urn:mpeg:dash:utc:http-xsdate:2014
urn:mpeg:dash:utc:http-iso:2014
urn:mpeg:dash:utc:http-ntp:2014
The manifest update task is used to poll the clock time server,
to save having to create a new thread.
When choosing the starting fragment number and when waiting for a
fragment to become available, the difference between the server's idea
of UTC and the client's idea of UTC is taken into account. For example,
if the server's time is behind the client's idea of UTC, we wait for
longer before requesting a fragment
[PDAM1]: http://www.iso.org/iso/home/store/catalogue_tc/catalogue_detail.htm?csnumber=66068
dashdemux: support NTP time servers in UTCTiming elements
Use the gst_ntp_clock to support the use of an NTP server.
https://bugzilla.gnome.org/show_bug.cgi?id=752413
This other type of baseURL test was replaced by a more complex one,
better have both to keep both options working
Also adds another 2 variations of how baseURL can be generated
https://bugzilla.gnome.org/show_bug.cgi?id=752776
gst_gl_memory_setup_wrapped() now takes a destroy notify function. This
destroy notify is called to track the memory life time, hence will
notify each time a memory get destroyed. This test check that the
callback count is correct.
We need to sync the pad values before taking the aggregator and pad locks
otherwise the element will just deadlock if there's any property changes
scheduled using GstController since that involves taking the aggregator and pad
locks.
Also add a test for this.
https://bugzilla.gnome.org/show_bug.cgi?id=749574
They require to get_proc_address some functions through the
platform specific {glX,egl}GetProcAddress rather than the default
GL library symbol lookup.
The previous approach of traversing the other_context weak ref tree was
1. Less performant
2. Incorrect for context destruction removing a link in the tree
Example of 2:
c1 = context_create (NULL)
c2 = context_create (c1)
c3 = context_create (c2)
context_can_share (c1, c3) == TRUE
context_destroy (c2)
unref (c2)
context_can_share (c1, c3) returns FALSE when it should be TRUE!
This does not remove the restriction that context sharedness can only
be tracked between GstGLContext's.
Added unit tests for all functions. Code coverage:
Overall coverage rate:
lines......: 83.8% (1941 of 2316 lines)
functions..: 100.0% (141 of 141 functions)
flexelint (http://www.gimpel.com/html/flex.htm) static code analyser
complained about implicit conversions from unsigned to signed, so I added
explicit conversions.
Ideally, the size parameter of gst_mpd_parse function should be unsigned,
but I don't want to change the API.
The duration_to_ms function converts a time specified by year, month, day,
hour, minute, second, millisecond to a millisecond value. Because all the
arguments are positive numbers, the result must also be positive.
This patch changes the returned value from a gint64 to a guint64 type.
Improved dash_mpd unit tests by adding new tests that parse the Period element.
Code coverage reported by lcov for dash/gstmpdparser.c is:
lines......: 43.0% (985 of 2290 lines)
functions..: 47.5% (67 of 141 functions)
According to ISO/IEC 23009-1:2014(E), chapter 5.3.2.1
"The Period extends until the PeriodStart of the next Period, or until
the end of the Media Presentation in the case of the last Period."
This means that a configured value for optional attribute period duration
should be ignored if the next period contains a start attribute or it is
the last period and the MPD contains a mediaPresentationDuration attribute.
https://bugzilla.gnome.org/show_bug.cgi?id=750797
The gst_mpdparser_get_rep_idx_with_max_bandwidth function assumes
representations are ordered by bandwidth and incorrectly returns the
first one when wanting the one with minimum bandwidth.
Corrected gst_mpdparser_get_rep_idx_with_max_bandwidth function to get the
correct representation in case max_bandwidth parameter is 0.
https://bugzilla.gnome.org/show_bug.cgi?id=751153
Added a check for a_node->ns before accessing a_node->ns->href in
gst_mpdparser_get_xml_node_namespace. This could happen if the xml
is missing the default namespace.
https://bugzilla.gnome.org/show_bug.cgi?id=750866
When the 'ignore-eos' property is set on a pad, compositor will keep resending
the last buffer on the pad till the pad is unlinked. We count the buffers
received on appsink, and if it's more than the buffers sent by videotestsrc, the
test passes.
Lets not cram everything into a single test - this would render the test name
useless for quick diagnosis. Having separate tests for the optional feature is
also verifying the behaviour when the feature is off.
Rather than one of the input pad video info's.
The test checking this was not constraining the output frame size
to ensure that the out of frame stream was not being displayed.
We verify that all the buffers on an obscured sinkpad are skipped by overriding
the map() function in the GstVideoMeta of the buffers to set a variable when
called. We also test that the buffers do get mapped when they're not obscured.
Blame^WCredit for the GstVideoMeta map() idea goes to Tim.
https://bugzilla.gnome.org/show_bug.cgi?id=746147
GstPhotography enables new paths in wrappercamerabinsrc that allows
the source to be notified about the capture caps and provide an
alternative caps if desired bypassing the negotiation (this doesn't
seem like a good idea these days). To make sure it keeps working
until we remove it from the API in favor of standard caps negotiation
features this test was added.
It adds 3 extra tests with a simple test source that will:
1) Test that capturing with ANY caps work
2) Test that capturing with a fixed caps work
3) Test that capturing with a fixed caps and having the source
pick a different resolution from GstPhotography API works
by having wrappercamerabinsrc crop the capture to the final
requested dimensions
A bitmask is 64 bits, but integer immediates are passed as int
in varargs, which happen to be 32 bit with high probability.
This triggered a valgrind jump-relies-on-uninitalized-value
report well away from the site, since it doesn't trigger on
stack accesses, and there must have been enough zeroes to stop
g_object_set at the right place.
The first output MPEG-TS packet that corresponds to a video input
buffer which had the delta flag cleared (i.e. was a keyframe)
should have the delta flag cleared as well.
This is needed e.g. by tcpserversink in order to keep track
of the last keyframe and be able to burst data to newly-
connected clients.
https://bugzilla.gnome.org/show_bug.cgi?id=706872
This reverts commit 1c77d12ce8.
"interlaced" in the caps don't mean the same thing as the SOF2 marker in the
JPEG format. This test passes because of broken behaviour.
The flush stop could have happened between the source trying
to push the segment event and the buffer, this would cause a warning.
Prevent that by taking the source's stream lock while flushing.
https://bugzilla.gnome.org/show_bug.cgi?id=742684
With the current audiomixer, the input caps need to be the same,
otherwise there is an unavoidable race in the caps negotiation. So
enforce that using capsfilters
https://bugzilla.gnome.org/show_bug.cgi?id=742684
Get rid of now-useless packetizer struct and just
call internal functions directly. Also remove
version property which is now defunct, not least
because we create the packetizer with the
version in the init function before a version
can be set.
When this is TRUE, we really have to produce output. This happens
in live mixing mode when we have to output something for the current
time, no matter if we have enough input or not.
In file included from /home/thiagoss/gst/head/gstreamer/gst/gst.h:54:0,
from /home/thiagoss/gst/head/gstreamer/libs/gst/check/gstcheck.h:34,
from elements/hlsdemux_m3u8.c:27:
../../ext/hls/gstfragmented.h:8:28: error: redundant redeclaration of ‘fragmented_debug’ [-Werror=redundant-decls]
GST_DEBUG_CATEGORY_EXTERN (fragmented_debug);
Move the definition of the category to after the declaration.
Ported from https://github.com/ylatuya/gst-plugins-bad
This still has some unit tests for alternative renditions and
seeking, which are commented out for the time being until we
support them properly.
Use the sticky events to compose the streamheader as they are the
ones that are persisted to config new pads linked. Instead of storing
them ourselves rely on the pad storage that already orders it for us
https://bugzilla.gnome.org/show_bug.cgi?id=732596
Check that end_of_seq() [EOSEQ] and end_of_stream [EOS] NAL units
are correctly parsed and the reported NAL unit size yields 1 byte,
i.e. the only NalHeaderBytes in there.
https://bugzilla.gnome.org/show_bug.cgi?id=732553
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
Check that conversion to byte-stream/au formats work and that we
can effectively drop broken/invalid NAL units from the resulting
access unit buffer.
https://bugzilla.gnome.org/show_bug.cgi?id=732203
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
If an SEI NAL unit with a buffering_period() message is inserted
between an SPS and PPS NAL unit, check that the output buffer still
contain it. i.e. make sure that this SEI message is not dropped.
https://bugzilla.gnome.org/show_bug.cgi?id=732156
Signed-off-by: Gwenole Beauchesne <gwenole.beauchesne@intel.com>
It was previously a mix and match of both variants, introducing just too much
confusion.
The prefix are from now on:
* GstMpegts for structures and type names (and not GstMpegTs)
* gst_mpegts_ for functions (and not gst_mpeg_ts_)
* GST_MPEGTS_ for enums/flags (and not GST_MPEG_TS_)
* GST_TYPE_MPEGTS_ for types (and not GST_TYPE_MPEG_TS_)
The rationale for chosing that is:
* the namespace is shorter/direct (it's mpegts, not mpeg_ts nor mpeg-ts)
* the namespace is one word under Gst
* it's shorter (yah)
Expose one more libcurl option: CURLOPT_SSH_HOST_PUBLIC_KEY_MD5.
This allows authenticating the server by the MD5 fingerprint of
the server's public key.
https://bugzilla.gnome.org/show_bug.cgi?id=723167
This patch provides the basic infrastructure required for this.
Upload and Download has been ported to this.
Has the nice effect of allowing GstGLMemory to be our
refcounted texture object for any texture type (not just RGBA).
Should not lose any features/video formats.
We create our textures (in Desktop GL) with GL_TEXTURE_RECTANGLE,
vaapi attempts to bind our texture to GL_TEXTURE_2D which throws a
GL_INVALID_OPERATION error and as thus, no video.
Also, by moving exclusively to GL_TEXTURE_2D and the npot extension
we also remove a difference between the Desktop GL and GLES2 code.
https://bugzilla.gnome.org/show_bug.cgi?id=712287
Tests for PAT, PMT, and NIT
Creates a new table, and populates it with descriptors.
Parses the newly created tables, and checks the data.
Creates a GstMpegTsSection from the tables, and packetize the sections.
The packetized section data is byte-wise compared to a static byte array
https://bugzilla.gnome.org/show_bug.cgi?id=723953
The thread that calls the success/failure callback can be the
same that is adding/removing the element as the IDLE probe can
happen instantly if the pad is not 'busy'.
This required moving some checks for the callback counter around
as well as removing some pad pushes from the main test thread as
they were made useless after the IDLE pad probe was fixed in core
by commit 0324358ebc
* stream-start-id is mandatory at the beginning, so add that to the
gdp headers
* caps must be sent before new segment, invert the order from legacy
0.10 code
And fix the tests as a ref is now kept for those buffers that compose
the header
Most of the tests weren't updated after the sticky events order
and stream start. Fix that and refactor those tests check that
are the same to some common functions.
Those functions still don't actually test the content but at
least now they are in a single place and can be improved
without replication
Commit 6af387cd5a made h264parse
strip a leading 0x00 byte from some output scenarios. This broke
tests as bs_to_nal test expects one more byte on the output.
Fix this by comparing the output with the expected stripped version,
too.
When outputting in AVC3 stream format, the codec_data should not
contain any SPS or PPS, because they are embedded inside the stream.
In case of avc->bytestream h264parse will push the SPS and PPS from
codec_data downstream at the start of the stream, at intervals
controlled by "config-interval" and when there is a codec_data change.
In the case of avc3->bytstream h264parse detects that there is
already SPS/PPS in the stream and sets h264parse->push_codec to FALSE.
Therefore avc3->bytstream was already supported, except for the stream
type.
In the case of bystream->avc h264parse will generate codec_data caps
from the parsed SPS/PPS in the stream. However it does not remove these
SPS/PPS from the stream. bytestream->avc3 is the same as bytestream->avc
except that the codec_data must not have any SPS/PPS in it.
|--------------+-------------+-------------------|
|stream-format | SPS in-band | SPS in codec_data |
|--------------+-------------+-------------------|
| avc | maybe | always |
|--------------+-------------+-------------------|
| avc3 | always | never |
|--------------+-------------+-------------------|
Amendment 2 of ISO/IEC 14496-15 (AVC file format) is defining a new
structure for fragmented MP4 called "avc3". The principal difference
between AVC1 and AVC3 is the location of the codec initialisation
data (e.g. SPS, PPS). In AVC1 this data is placed in the initial MOOV box
(moov.trak.mdia.minf.stbl.stsd.avc1) but in AVC3 this data goes in the
first sample of every fragment.
https://bugzilla.gnome.org/show_bug.cgi?id=702004
Do state changes from sink to src. Fixes race condition in
pull mode test where the source will start up and push buffers
to queue/identity or aiffparse before the main thread has
managed to set them to playing yet.
